Open data: global wealth distribution by band from the Credit Suisse and UBS Global Wealth Reports, 2010 to 2023 (Global)

Rows: 44
Columns: 8

Wealth shares across four wealth bands for the global adult population, plus mean, median, and Gini figures for 20 countries. Free to download at datahub.io. Widely used in inequality research and public communication.

AI analysis

Dataset Analysis: Global Wealth Distribution (Credit Suisse / UBS, 2010–2023)

1. Dataset Overview & Research Value

This dataset captures global wealth distribution across four discrete wealth bands for the adult population worldwide, spanning 14 years (2010–2023) with 11 distinct time points, drawn from the Credit Suisse and UBS Global Wealth Reports — one of the most cited annual wealth benchmarking series in academic and policy research. The dataset's availability on datahub.io signals it is intended for broad public use, and its coverage through 2023 makes it reasonably current. Its core value lies in tracking how wealth concentration has shifted over time across bottom, middle, upper-middle, and top wealth tiers, making it well-suited for longitudinal inequality analysis. With only 44 rows, the dataset is compact but dense with meaning — each row represents a wealth band × year combination, enabling trend and share-shift studies without large-scale data infrastructure.

2. Data Quality Observations

Overall data quality is good but not perfect. There are zero null values across six of eight columns, which is commendable for a curated research dataset. However, upperboundusd carries 11 null values — exactly 25% of all rows — which is structurally significant rather than accidental: the top wealth band ("$1M") has no finite upper bound by definition, so these nulls are semantically valid but should be explicitly documented as right-open intervals rather than missing data. Analysts must handle this carefully to avoid inadvertently dropping the wealthiest tier in any filtering or join operation. No type mismatches are flagged, and with 0 duplicate rows reported the dataset appears clean and deduplicated. The one structural caution is that year has only 11 distinct values across 44 rows, confirming the 4-band × 11-year panel structure — but analysts should verify whether any years are missing mid-series, since a 2010–2023 span would suggest up to 14 annual observations, implying 3 years may be absent from the record.

3. Key Column Distributions & Statistical Highlights

The most analytically rich columns are adultsmillions, wealthusdbillions, and wealthsharepct. The adultsmillions column ranges from 24M to 3,474M with a mean of 1,241M and median of 745M — a substantial right skew (0.638) reflecting that the lower wealth bands contain the vast majority of the global adult population. Similarly, wealthusdbillions spans $5T to $208.3T with a mean of $73T and median of $56T (skew=0.785), underscoring heavy concentration at the top band. The wealthsharepct column, by contrast, is the most normally distributed of the group (skew=0.402), ranging from 1.1% to 62.7% — likely because proportional shares are mathematically constrained to sum to 100% per year, moderating extremes. On the bounds side, lowerboundusd shows high skewness (1.17) with 11 flagged outliers, driven entirely by the $1M threshold of the top band dwarfing the others ($0, $10K, $100K) — again a structural feature of wealth band design rather than data error. The adultssharepct column mirrors the population distribution pattern (range: 0.5%–73.2%), reinforcing that the sub-$10K band historically captures the majority of adults globally.

4. Recommended Join Key Columns

The natural composite join key is year + wealthband, which together uniquely identify each of the 44 rows. For cross-dataset joins, year (as a Date/integer) is the primary temporal anchor and should be standardized to a 4-digit year integer for compatibility with most macroeconomic datasets. The wealthband string field (4 distinct values) serves as a categorical dimension key — though analysts should anticipate label normalization challenges when joining to other inequality datasets that may use different band naming conventions (e.g., decile-based vs. threshold-based classifications). If country-level extensions are added (the description mentions mean, median, and Gini figures for 20 countries), a country key would become essential. For now, all records represent global aggregates, so joins should be scoped accordingly.

5. Recommended Pairing Datasets for Correlation Discovery

Several dataset categories would pair powerfully with this data:

- World Bank GDP & GNI per capita series — correlating wealth band shifts with global economic growth cycles to test whether expansions disproportionately benefit top-tier holders - IMF or OECD Gini coefficient datasets — the description notes Gini figures exist for 20 countries; joining external Gini time series would enable validation and richer inequality trend analysis - Global financial asset price indices (equity markets, real estate) — since upper-band wealth is heavily asset-driven, S&P 500 or MSCI World returns by year could explain variance in top-band wealth share - Tax policy and fiscal data (e.g., OECD Tax Database, Tax Foundation) — to explore whether top marginal rates or capital gains tax changes correlate with wealth concentration shifts - UN Human Development Index (HDI) or poverty headcount data — to examine whether growth in the sub-$10K band corresponds to improvements in development metrics, testing "trickle-down" hypotheses empirically

Columns

year (date)
wealth_band (string)
lower_bound_usd (integer)
upper_bound_usd (decimal)
adults_millions (decimal)
adults_share_pct (decimal)
wealth_usd_billions (integer)
wealth_share_pct (decimal)